OCT Segmentation Errors with Bruch's Membrane Opening-Minimum Rim Width as Compared with Retinal Nerve Fiber Layer...

Area of Science:

• Ophthalmology and Optometry
• Medical Imaging Analysis

Background:

• Accurate measurement of optic nerve head parameters is crucial for glaucoma diagnosis and management.
• Automated segmentation algorithms in optical coherence tomography (OCT) are widely used but prone to errors.

Purpose of the Study:

• To compare the magnitude and location of automated segmentation errors between Bruch's membrane opening-minimum rim width (BMO-MRW) and retinal nerve fiber layer thickness (RNFLT) measurements.
• To assess the impact of these errors on glaucoma classification.

Main Methods:

• Cross-sectional study including 162 eyes of glaucoma suspects or patients with open-angle glaucoma.
• Spectral-domain OCT imaging was performed, and automated segmentation results for BMO-MRW and RNFLT were exported.
• Absolute and proportional errors were calculated globally and by sector, and glaucoma classification differences were analyzed.

Main Results:

• Absolute segmentation error was significantly larger for BMO-MRW (10.8 μm) than for RNFLT (3.58 μm), but proportional errors were similar (4.3% vs. 4.4%).
• RNFLT errors were associated with specific sectors (superior/inferior), thicker nerve fiber layers, and worse visual fields.
• BMO-MRW and RNFLT errors did not commonly occur in the same sector, and manual refinement changed glaucoma classification in 6-8% of eyes.

Conclusions:

• Both BMO-MRW and RNFLT measurements exhibit segmentation errors.
• These errors do not consistently overlap in location and can potentially influence glaucoma classification.
• Understanding segmentation error patterns is important for accurate glaucoma assessment using OCT.